1 /*
2 * Copyright (C) 2001 Sistina Software (UK) Limited.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
4 *
5 * This file is released under the LGPL.
6 */
7
8 #ifndef _LINUX_DEVICE_MAPPER_H
9 #define _LINUX_DEVICE_MAPPER_H
10
11 #include <linux/bio.h>
12 #include <linux/blkdev.h>
13 #include <linux/dm-ioctl.h>
14 #include <linux/math64.h>
15 #include <linux/ratelimit.h>
16
17 struct dm_dev;
18 struct dm_target;
19 struct dm_table;
20 struct dm_report_zones_args;
21 struct mapped_device;
22 struct bio_vec;
23 enum dax_access_mode;
24
25 /*
26 * Type of table, mapped_device's mempool and request_queue
27 */
28 enum dm_queue_mode {
29 DM_TYPE_NONE = 0,
30 DM_TYPE_BIO_BASED = 1,
31 DM_TYPE_REQUEST_BASED = 2,
32 DM_TYPE_DAX_BIO_BASED = 3,
33 };
34
35 typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE, STATUSTYPE_IMA } status_type_t;
36
37 union map_info {
38 void *ptr;
39 };
40
41 /*
42 * In the constructor the target parameter will already have the
43 * table, type, begin and len fields filled in.
44 */
45 typedef int (*dm_ctr_fn) (struct dm_target *target,
46 unsigned int argc, char **argv);
47
48 /*
49 * The destructor doesn't need to free the dm_target, just
50 * anything hidden ti->private.
51 */
52 typedef void (*dm_dtr_fn) (struct dm_target *ti);
53
54 /*
55 * The map function must return:
56 * < 0: error
57 * = 0: The target will handle the io by resubmitting it later
58 * = 1: simple remap complete
59 * = 2: The target wants to push back the io
60 */
61 typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio);
62 typedef int (*dm_clone_and_map_request_fn) (struct dm_target *ti,
63 struct request *rq,
64 union map_info *map_context,
65 struct request **clone);
66 typedef void (*dm_release_clone_request_fn) (struct request *clone,
67 union map_info *map_context);
68
69 /*
70 * Returns:
71 * < 0 : error (currently ignored)
72 * 0 : ended successfully
73 * 1 : for some reason the io has still not completed (eg,
74 * multipath target might want to requeue a failed io).
75 * 2 : The target wants to push back the io
76 */
77 typedef int (*dm_endio_fn) (struct dm_target *ti,
78 struct bio *bio, blk_status_t *error);
79 typedef int (*dm_request_endio_fn) (struct dm_target *ti,
80 struct request *clone, blk_status_t error,
81 union map_info *map_context);
82
83 typedef void (*dm_presuspend_fn) (struct dm_target *ti);
84 typedef void (*dm_presuspend_undo_fn) (struct dm_target *ti);
85 typedef void (*dm_postsuspend_fn) (struct dm_target *ti);
86 typedef int (*dm_preresume_fn) (struct dm_target *ti);
87 typedef void (*dm_resume_fn) (struct dm_target *ti);
88
89 typedef void (*dm_status_fn) (struct dm_target *ti, status_type_t status_type,
90 unsigned status_flags, char *result, unsigned maxlen);
91
92 typedef int (*dm_message_fn) (struct dm_target *ti, unsigned argc, char **argv,
93 char *result, unsigned maxlen);
94
95 typedef int (*dm_prepare_ioctl_fn) (struct dm_target *ti, struct block_device **bdev);
96
97 #ifdef CONFIG_BLK_DEV_ZONED
98 typedef int (*dm_report_zones_fn) (struct dm_target *ti,
99 struct dm_report_zones_args *args,
100 unsigned int nr_zones);
101 #else
102 /*
103 * Define dm_report_zones_fn so that targets can assign to NULL if
104 * CONFIG_BLK_DEV_ZONED disabled. Otherwise each target needs to do
105 * awkward #ifdefs in their target_type, etc.
106 */
107 typedef int (*dm_report_zones_fn) (struct dm_target *dummy);
108 #endif
109
110 /*
111 * These iteration functions are typically used to check (and combine)
112 * properties of underlying devices.
113 * E.g. Does at least one underlying device support flush?
114 * Does any underlying device not support WRITE_SAME?
115 *
116 * The callout function is called once for each contiguous section of
117 * an underlying device. State can be maintained in *data.
118 * Return non-zero to stop iterating through any further devices.
119 */
120 typedef int (*iterate_devices_callout_fn) (struct dm_target *ti,
121 struct dm_dev *dev,
122 sector_t start, sector_t len,
123 void *data);
124
125 /*
126 * This function must iterate through each section of device used by the
127 * target until it encounters a non-zero return code, which it then returns.
128 * Returns zero if no callout returned non-zero.
129 */
130 typedef int (*dm_iterate_devices_fn) (struct dm_target *ti,
131 iterate_devices_callout_fn fn,
132 void *data);
133
134 typedef void (*dm_io_hints_fn) (struct dm_target *ti,
135 struct queue_limits *limits);
136
137 /*
138 * Returns:
139 * 0: The target can handle the next I/O immediately.
140 * 1: The target can't handle the next I/O immediately.
141 */
142 typedef int (*dm_busy_fn) (struct dm_target *ti);
143
144 /*
145 * Returns:
146 * < 0 : error
147 * >= 0 : the number of bytes accessible at the address
148 */
149 typedef long (*dm_dax_direct_access_fn) (struct dm_target *ti, pgoff_t pgoff,
150 long nr_pages, enum dax_access_mode node, void **kaddr,
151 pfn_t *pfn);
152 typedef int (*dm_dax_zero_page_range_fn)(struct dm_target *ti, pgoff_t pgoff,
153 size_t nr_pages);
154
155 /*
156 * Returns:
157 * != 0 : number of bytes transferred
158 * 0 : recovery write failed
159 */
160 typedef size_t (*dm_dax_recovery_write_fn)(struct dm_target *ti, pgoff_t pgoff,
161 void *addr, size_t bytes, struct iov_iter *i);
162
163 void dm_error(const char *message);
164
165 struct dm_dev {
166 struct block_device *bdev;
167 struct dax_device *dax_dev;
168 fmode_t mode;
169 char name[16];
170 };
171
172 dev_t dm_get_dev_t(const char *path);
173
174 /*
175 * Constructors should call these functions to ensure destination devices
176 * are opened/closed correctly.
177 */
178 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
179 struct dm_dev **result);
180 void dm_put_device(struct dm_target *ti, struct dm_dev *d);
181
182 /*
183 * Information about a target type
184 */
185
186 struct target_type {
187 uint64_t features;
188 const char *name;
189 struct module *module;
190 unsigned version[3];
191 dm_ctr_fn ctr;
192 dm_dtr_fn dtr;
193 dm_map_fn map;
194 dm_clone_and_map_request_fn clone_and_map_rq;
195 dm_release_clone_request_fn release_clone_rq;
196 dm_endio_fn end_io;
197 dm_request_endio_fn rq_end_io;
198 dm_presuspend_fn presuspend;
199 dm_presuspend_undo_fn presuspend_undo;
200 dm_postsuspend_fn postsuspend;
201 dm_preresume_fn preresume;
202 dm_resume_fn resume;
203 dm_status_fn status;
204 dm_message_fn message;
205 dm_prepare_ioctl_fn prepare_ioctl;
206 dm_report_zones_fn report_zones;
207 dm_busy_fn busy;
208 dm_iterate_devices_fn iterate_devices;
209 dm_io_hints_fn io_hints;
210 dm_dax_direct_access_fn direct_access;
211 dm_dax_zero_page_range_fn dax_zero_page_range;
212 dm_dax_recovery_write_fn dax_recovery_write;
213
214 /* For internal device-mapper use. */
215 struct list_head list;
216 };
217
218 /*
219 * Target features
220 */
221
222 /*
223 * Any table that contains an instance of this target must have only one.
224 */
225 #define DM_TARGET_SINGLETON 0x00000001
226 #define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON)
227
228 /*
229 * Indicates that a target does not support read-only devices.
230 */
231 #define DM_TARGET_ALWAYS_WRITEABLE 0x00000002
232 #define dm_target_always_writeable(type) \
233 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE)
234
235 /*
236 * Any device that contains a table with an instance of this target may never
237 * have tables containing any different target type.
238 */
239 #define DM_TARGET_IMMUTABLE 0x00000004
240 #define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE)
241
242 /*
243 * Indicates that a target may replace any target; even immutable targets.
244 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined.
245 */
246 #define DM_TARGET_WILDCARD 0x00000008
247 #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD)
248
249 /*
250 * A target implements own bio data integrity.
251 */
252 #define DM_TARGET_INTEGRITY 0x00000010
253 #define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY)
254
255 /*
256 * A target passes integrity data to the lower device.
257 */
258 #define DM_TARGET_PASSES_INTEGRITY 0x00000020
259 #define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY)
260
261 /*
262 * Indicates support for zoned block devices:
263 * - DM_TARGET_ZONED_HM: the target also supports host-managed zoned
264 * block devices but does not support combining different zoned models.
265 * - DM_TARGET_MIXED_ZONED_MODEL: the target supports combining multiple
266 * devices with different zoned models.
267 */
268 #ifdef CONFIG_BLK_DEV_ZONED
269 #define DM_TARGET_ZONED_HM 0x00000040
270 #define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM)
271 #else
272 #define DM_TARGET_ZONED_HM 0x00000000
273 #define dm_target_supports_zoned_hm(type) (false)
274 #endif
275
276 /*
277 * A target handles REQ_NOWAIT
278 */
279 #define DM_TARGET_NOWAIT 0x00000080
280 #define dm_target_supports_nowait(type) ((type)->features & DM_TARGET_NOWAIT)
281
282 /*
283 * A target supports passing through inline crypto support.
284 */
285 #define DM_TARGET_PASSES_CRYPTO 0x00000100
286 #define dm_target_passes_crypto(type) ((type)->features & DM_TARGET_PASSES_CRYPTO)
287
288 #ifdef CONFIG_BLK_DEV_ZONED
289 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000200
290 #define dm_target_supports_mixed_zoned_model(type) \
291 ((type)->features & DM_TARGET_MIXED_ZONED_MODEL)
292 #else
293 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000000
294 #define dm_target_supports_mixed_zoned_model(type) (false)
295 #endif
296
297 struct dm_target {
298 struct dm_table *table;
299 struct target_type *type;
300
301 /* target limits */
302 sector_t begin;
303 sector_t len;
304
305 /* If non-zero, maximum size of I/O submitted to a target. */
306 uint32_t max_io_len;
307
308 /*
309 * A number of zero-length barrier bios that will be submitted
310 * to the target for the purpose of flushing cache.
311 *
312 * The bio number can be accessed with dm_bio_get_target_bio_nr.
313 * It is a responsibility of the target driver to remap these bios
314 * to the real underlying devices.
315 */
316 unsigned num_flush_bios;
317
318 /*
319 * The number of discard bios that will be submitted to the target.
320 * The bio number can be accessed with dm_bio_get_target_bio_nr.
321 */
322 unsigned num_discard_bios;
323
324 /*
325 * The number of secure erase bios that will be submitted to the target.
326 * The bio number can be accessed with dm_bio_get_target_bio_nr.
327 */
328 unsigned num_secure_erase_bios;
329
330 /*
331 * The number of WRITE ZEROES bios that will be submitted to the target.
332 * The bio number can be accessed with dm_bio_get_target_bio_nr.
333 */
334 unsigned num_write_zeroes_bios;
335
336 /*
337 * The minimum number of extra bytes allocated in each io for the
338 * target to use.
339 */
340 unsigned per_io_data_size;
341
342 /* target specific data */
343 void *private;
344
345 /* Used to provide an error string from the ctr */
346 char *error;
347
348 /*
349 * Set if this target needs to receive flushes regardless of
350 * whether or not its underlying devices have support.
351 */
352 bool flush_supported:1;
353
354 /*
355 * Set if this target needs to receive discards regardless of
356 * whether or not its underlying devices have support.
357 */
358 bool discards_supported:1;
359
360 /*
361 * Set if we need to limit the number of in-flight bios when swapping.
362 */
363 bool limit_swap_bios:1;
364
365 /*
366 * Set if this target implements a zoned device and needs emulation of
367 * zone append operations using regular writes.
368 */
369 bool emulate_zone_append:1;
370
371 /*
372 * Set if the target will submit IO using dm_submit_bio_remap()
373 * after returning DM_MAPIO_SUBMITTED from its map function.
374 */
375 bool accounts_remapped_io:1;
376
377 /*
378 * Set if the target will submit the DM bio without first calling
379 * bio_set_dev(). NOTE: ideally a target should _not_ need this.
380 */
381 bool needs_bio_set_dev:1;
382 };
383
384 void *dm_per_bio_data(struct bio *bio, size_t data_size);
385 struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size);
386 unsigned dm_bio_get_target_bio_nr(const struct bio *bio);
387
388 u64 dm_start_time_ns_from_clone(struct bio *bio);
389
390 int dm_register_target(struct target_type *t);
391 void dm_unregister_target(struct target_type *t);
392
393 /*
394 * Target argument parsing.
395 */
396 struct dm_arg_set {
397 unsigned argc;
398 char **argv;
399 };
400
401 /*
402 * The minimum and maximum value of a numeric argument, together with
403 * the error message to use if the number is found to be outside that range.
404 */
405 struct dm_arg {
406 unsigned min;
407 unsigned max;
408 char *error;
409 };
410
411 /*
412 * Validate the next argument, either returning it as *value or, if invalid,
413 * returning -EINVAL and setting *error.
414 */
415 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
416 unsigned *value, char **error);
417
418 /*
419 * Process the next argument as the start of a group containing between
420 * arg->min and arg->max further arguments. Either return the size as
421 * *num_args or, if invalid, return -EINVAL and set *error.
422 */
423 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
424 unsigned *num_args, char **error);
425
426 /*
427 * Return the current argument and shift to the next.
428 */
429 const char *dm_shift_arg(struct dm_arg_set *as);
430
431 /*
432 * Move through num_args arguments.
433 */
434 void dm_consume_args(struct dm_arg_set *as, unsigned num_args);
435
436 /*-----------------------------------------------------------------
437 * Functions for creating and manipulating mapped devices.
438 * Drop the reference with dm_put when you finish with the object.
439 *---------------------------------------------------------------*/
440
441 /*
442 * DM_ANY_MINOR chooses the next available minor number.
443 */
444 #define DM_ANY_MINOR (-1)
445 int dm_create(int minor, struct mapped_device **md);
446
447 /*
448 * Reference counting for md.
449 */
450 struct mapped_device *dm_get_md(dev_t dev);
451 void dm_get(struct mapped_device *md);
452 int dm_hold(struct mapped_device *md);
453 void dm_put(struct mapped_device *md);
454
455 /*
456 * An arbitrary pointer may be stored alongside a mapped device.
457 */
458 void dm_set_mdptr(struct mapped_device *md, void *ptr);
459 void *dm_get_mdptr(struct mapped_device *md);
460
461 /*
462 * A device can still be used while suspended, but I/O is deferred.
463 */
464 int dm_suspend(struct mapped_device *md, unsigned suspend_flags);
465 int dm_resume(struct mapped_device *md);
466
467 /*
468 * Event functions.
469 */
470 uint32_t dm_get_event_nr(struct mapped_device *md);
471 int dm_wait_event(struct mapped_device *md, int event_nr);
472 uint32_t dm_next_uevent_seq(struct mapped_device *md);
473 void dm_uevent_add(struct mapped_device *md, struct list_head *elist);
474
475 /*
476 * Info functions.
477 */
478 const char *dm_device_name(struct mapped_device *md);
479 int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid);
480 struct gendisk *dm_disk(struct mapped_device *md);
481 int dm_suspended(struct dm_target *ti);
482 int dm_post_suspending(struct dm_target *ti);
483 int dm_noflush_suspending(struct dm_target *ti);
484 void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors);
485 void dm_submit_bio_remap(struct bio *clone, struct bio *tgt_clone);
486 union map_info *dm_get_rq_mapinfo(struct request *rq);
487
488 #ifdef CONFIG_BLK_DEV_ZONED
489 struct dm_report_zones_args {
490 struct dm_target *tgt;
491 sector_t next_sector;
492
493 void *orig_data;
494 report_zones_cb orig_cb;
495 unsigned int zone_idx;
496
497 /* must be filled by ->report_zones before calling dm_report_zones_cb */
498 sector_t start;
499 };
500 int dm_report_zones(struct block_device *bdev, sector_t start, sector_t sector,
501 struct dm_report_zones_args *args, unsigned int nr_zones);
502 #endif /* CONFIG_BLK_DEV_ZONED */
503
504 /*
505 * Device mapper functions to parse and create devices specified by the
506 * parameter "dm-mod.create="
507 */
508 int __init dm_early_create(struct dm_ioctl *dmi,
509 struct dm_target_spec **spec_array,
510 char **target_params_array);
511
512 struct queue_limits *dm_get_queue_limits(struct mapped_device *md);
513
514 /*
515 * Geometry functions.
516 */
517 int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo);
518 int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo);
519
520 /*-----------------------------------------------------------------
521 * Functions for manipulating device-mapper tables.
522 *---------------------------------------------------------------*/
523
524 /*
525 * First create an empty table.
526 */
527 int dm_table_create(struct dm_table **result, fmode_t mode,
528 unsigned num_targets, struct mapped_device *md);
529
530 /*
531 * Then call this once for each target.
532 */
533 int dm_table_add_target(struct dm_table *t, const char *type,
534 sector_t start, sector_t len, char *params);
535
536 /*
537 * Target can use this to set the table's type.
538 * Can only ever be called from a target's ctr.
539 * Useful for "hybrid" target (supports both bio-based
540 * and request-based).
541 */
542 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type);
543
544 /*
545 * Finally call this to make the table ready for use.
546 */
547 int dm_table_complete(struct dm_table *t);
548
549 /*
550 * Destroy the table when finished.
551 */
552 void dm_table_destroy(struct dm_table *t);
553
554 /*
555 * Target may require that it is never sent I/O larger than len.
556 */
557 int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len);
558
559 /*
560 * Table reference counting.
561 */
562 struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx);
563 void dm_put_live_table(struct mapped_device *md, int srcu_idx);
564 void dm_sync_table(struct mapped_device *md);
565
566 /*
567 * Queries
568 */
569 sector_t dm_table_get_size(struct dm_table *t);
570 unsigned int dm_table_get_num_targets(struct dm_table *t);
571 fmode_t dm_table_get_mode(struct dm_table *t);
572 struct mapped_device *dm_table_get_md(struct dm_table *t);
573 const char *dm_table_device_name(struct dm_table *t);
574
575 /*
576 * Trigger an event.
577 */
578 void dm_table_event(struct dm_table *t);
579
580 /*
581 * Run the queue for request-based targets.
582 */
583 void dm_table_run_md_queue_async(struct dm_table *t);
584
585 /*
586 * The device must be suspended before calling this method.
587 * Returns the previous table, which the caller must destroy.
588 */
589 struct dm_table *dm_swap_table(struct mapped_device *md,
590 struct dm_table *t);
591
592 /*
593 * Table blk_crypto_profile functions
594 */
595 void dm_destroy_crypto_profile(struct blk_crypto_profile *profile);
596
597 /*-----------------------------------------------------------------
598 * Macros.
599 *---------------------------------------------------------------*/
600 #define DM_NAME "device-mapper"
601
602 #define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n"
603
604 #define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__)
605
606 #define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__)
607 #define DMERR_LIMIT(fmt, ...) pr_err_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
608 #define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__)
609 #define DMWARN_LIMIT(fmt, ...) pr_warn_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
610 #define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__)
611 #define DMINFO_LIMIT(fmt, ...) pr_info_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
612
613 #define DMDEBUG(fmt, ...) pr_debug(DM_FMT(fmt), ##__VA_ARGS__)
614 #define DMDEBUG_LIMIT(fmt, ...) pr_debug_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
615
616 #define DMEMIT(x...) sz += ((sz >= maxlen) ? \
617 0 : scnprintf(result + sz, maxlen - sz, x))
618
619 #define DMEMIT_TARGET_NAME_VERSION(y) \
620 DMEMIT("target_name=%s,target_version=%u.%u.%u", \
621 (y)->name, (y)->version[0], (y)->version[1], (y)->version[2])
622
623 /*
624 * Definitions of return values from target end_io function.
625 */
626 #define DM_ENDIO_DONE 0
627 #define DM_ENDIO_INCOMPLETE 1
628 #define DM_ENDIO_REQUEUE 2
629 #define DM_ENDIO_DELAY_REQUEUE 3
630
631 /*
632 * Definitions of return values from target map function.
633 */
634 #define DM_MAPIO_SUBMITTED 0
635 #define DM_MAPIO_REMAPPED 1
636 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE
637 #define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE
638 #define DM_MAPIO_KILL 4
639
640 #define dm_sector_div64(x, y)( \
641 { \
642 u64 _res; \
643 (x) = div64_u64_rem(x, y, &_res); \
644 _res; \
645 } \
646 )
647
648 /*
649 * Ceiling(n / sz)
650 */
651 #define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz))
652
653 #define dm_sector_div_up(n, sz) ( \
654 { \
655 sector_t _r = ((n) + (sz) - 1); \
656 sector_div(_r, (sz)); \
657 _r; \
658 } \
659 )
660
661 /*
662 * ceiling(n / size) * size
663 */
664 #define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz))
665
666 /*
667 * Sector offset taken relative to the start of the target instead of
668 * relative to the start of the device.
669 */
670 #define dm_target_offset(ti, sector) ((sector) - (ti)->begin)
671
to_sector(unsigned long long n)672 static inline sector_t to_sector(unsigned long long n)
673 {
674 return (n >> SECTOR_SHIFT);
675 }
676
to_bytes(sector_t n)677 static inline unsigned long to_bytes(sector_t n)
678 {
679 return (n << SECTOR_SHIFT);
680 }
681
682 #endif /* _LINUX_DEVICE_MAPPER_H */
683